#149 Designing AI Features: The 3-Step Framework Product Teams Need with Bansi Mehta

👉 Join the Case Study Engine waitlist: https://ai-for-designers.com/claudecodeportfolio-waitlist The waitlist opens first on 9 November, gets the lowest price and an exclusive live Q&A call with me. A client asked me for my projects, and I had nothing written up. This quarterly starts right there. In this episode: Claude Motion, and why it isn’t really a video tool Agents and skills inside Figma, and why the most powerful part is a text file Cheaper Claude models (Opus 5.5, Sonnet 5.5, Haiku 5.5) and what that means for building your own tools Automation that keeps running in the background Design moving into Claude Code Then what I built from that client moment, the Case Study Engine. You put everything about a project in once, and it builds an honest, visual case study for recruiters and clients. Plus what I learned in week three of the AI for Designers bootcamp, what didn’t go to plan, and where I think our job is heading.

Meta's Muse is one of the best designed AI products of the year, and its little keychain, the Muse Charm, is coming by December. In this episode I look at what is inside the horse: hourly files on you and the people you mention, training switched on by default, a test with human callers, and an agent that couldn't explain what it had access to. Then I put Muse next to Meta's history of gifts first and terms later, from Onavo to the Ray-Ban glasses to our Instagram photos. I argue against myself, and I finish with three design decisions and three predictions on where consent goes once the screen disappears.

This summer, more than a thousand AI agents broke out of a test at OpenAI and hacked Hugging Face, because they were looking for the answers to an exam. A few weeks later a 27 year old researcher quit his job at Anthropic and said the AI labs are gambling with our lives, and since then it feels like everybody is afraid of AI. In this episode I tell you the whole story from the beginning. What happened in that test lab, who Jacob Coxon is, why some people want to stop AI development and why others say the fear is hype or even marketing. And at the end I share what I think it means for us as designers, because almost everything that went wrong here is something we work on every day.

👉 Grab one of the last seat for AI for Designers: https://ai-for-designers.com AI is taking over more and more of the making. What's left for us as designers is the meeting where it all gets decided, and I think most of us never learned how to handle that part. In this episode I talk with facilitator and workshop designer Brittni Bowering, who has run sessions for teams at Google, Procter & Gamble, Lufthansa and PVH. We skip the career story and go straight into three meetings you have probably sat in over the last two weeks. First the kickoff where the PM shows up with a spec from ChatGPT and says the thinking is done. Then the design review where three people vibe coded their own prototype and everyone argues about which one looks better. And then the research readout with synthetic users whose quotes are a little too perfect. For each one you get the actual words: the first question to ask, how to challenge the process without attacking the person, how to get a room to a decision with clear criteria and a quick vote, and what to say when your gut tells you the research is off but you have no data to prove it. Brittni also shares the three things she brings into every meeting, what stand-up comedy taught her about reading a room, and why the IKEA effect is the best way to stop being seen as the blocker. Brittni is also joining the AI for Designers Bootcamp as a guest speaker this October.

🔵 Open for enrollment: AI for Designers Early Bird Until Thursday: ⁠https://ai-for-designers.com⁠ A special episode, a little outside the usual rhythm. Doors for the next round of my AI for Designers Bootcamp opened today. The early bird runs until Thursday, and there are only a few spots left. If you're still on the fence, this one is for you. I'll tell you what happens inside the bootcamp and how a Monday there feels. You'll hear why it's built the way it is, starting with the moment earlier this year when I got stuck myself trying to automate one small task. We walk through all five weeks: research and prompting, AI prototyping, content generation with Adobe Firefly, UX for AI products, and the future of AI and design, where everyone presents their own project. You'll also hear how the design systems session with Cristian Morales Achiardi almost didn't happen. I'll share what came up in last Thursday's clinic, when one designer sent in her whole AI setup for review. And I'll be clear about who this bootcamp isn't for. We finish with what I think this means for us as designers over the next two years, and why none of us should be learning this alone.